The impact of phage predation on bacterial pathogens in the context of human disease is not currently appreciated. Here, we show that predatory interactions of a phage with an important environmentally transmitted pathogen, Vibrio cholerae, can modulate the evolutionary trajectory of this pathogen during the natural course of infection within individual patients. We analyzed geographically and temporally disparate cholera patient stool samples from Haiti and Bangladesh and found that phage predation can drive the genomic diversity of intra-patient V. cholerae populations. Intra-patient phage-sensitive and phage-resistant isolates were isogenic except for mutations conferring phage resistance, and moreover, phage-resistant V. cholerae populations were composed of a heterogeneous mix of many unique mutants. We also observed that phage predation can significantly alter the virulence potential of V. cholerae shed from cholera patients. We provide the first molecular evidence for predatory phage shaping microbial community structure during the natural course of infection in humans.https://doi.org/10.7554/eLife.03497.001
Cholera epidemics occur seasonally in areas such as Bangladesh, and outbreaks can also strike in vulnerable regions, as has occurred recently in Haiti. The disease is caused by Vibrio cholerae, a water-borne bacterium that colonizes the small intestine, and its symptoms include severe diarrhea and vomiting which can lead to death if the patient is not treated promptly.
Lytic phages are viruses that specifically attack and kill bacteria. After replicating many times inside the bacterial cell, the phages break open and destroy the cell. Over time a bacterial population can evolve to resist this phage ‘predation’; however, it is not known if bacterial pathogens need to defend themselves against phage attack when they infect humans. It had been suggested that phages might affect the progress of cholera infections in people, but molecular evidence that supports this hypothesis was lacking.
When testing stool samples from Haitian cholera patients, Seed et al. found one sample contained a lot of lytic phage relative to the amount of V. cholerae present. This phage was very similar to—but distinct from—a phage found in Bangladeshi patients.
The V. cholerae bacteria isolated from the stool sample were resistant to attack by the phage. Sequencing the genome of individual bacteria from this sample revealed that each had a mutation that made them resistant to the phage; and while many types of these mutations were found, these were the only differences between all the V. cholerae bacteria in this patient sample. This suggests that this resistance developed independently many different times within the patient due to strong selective pressure from phage predation.
When Seed et al. looked at a phage-positive stool sample from a Bangladeshi patient, more mutations that made the bacteria resistant to this phage were found; however, these mutations were different again from the ones in the Haitian bacteria. Because of the nature of these mutations the bacteria from this patient were rendered unable to cause disease and non-transmissible.
This work shows that phages can indeed have access to pathogenic bacteria during human infection. It also indicates that the pressure imposed by phage predation can, in some cases, be so strong that the bacteria lose their virulence and ability to spread to other humans in order to become resistant to the phage.https://doi.org/10.7554/eLife.03497.002
Traditional views of host–pathogen interactions are of a battle between two opposing organisms. This perspective is being challenged with greater appreciation of the influence of the host's microbial ecosystem on these interactions (Lozupone et al., 2012). Bacteriophages influence bacterial populations in many ecosystems and specifically temperate phages play a role in many diseases through lysogenic conversion (Brüssow et al., 2004). In contrast, the impact of lytic phage predation on bacterial pathogens in the context of human disease is under-appreciated. Vibrio cholerae is a water-borne pathogen responsible for the diarrheal disease cholera. In order for V. cholerae to colonize the human small intestine and elicit diarrhea, it produces a set of virulence determinants including cholera toxin (Herrington et al., 1988). The ToxR signaling cascade activates expression of these virulence factors in response to host environmental stimuli (Miller et al., 1987; Taylor et al., 1987; DiRita et al., 1991; Childers and Klose, 2007). Cholera epidemics appear with regular seasonality in regions like Bangladesh (Faruque et al., 2005) and can arise unpredictably in vulnerable regions as exemplified by the epidemic that recently began in Haiti following the single-source introduction of a pandemic V. cholerae O1 strain from another continent (Chin et al., 2011; Cravioto et al., 2011; Frerichs et al., 2012; Katz et al., 2013). Lytic phages are hypothesized to impact cholera disease burden in Bangladesh (Faruque et al., 2005). The predation of V. cholerae by phages following their co-ingestion from the environment is central to this hypothesis; however, molecular evidence in support of this hypothesis is lacking.
We recently described the ICP2 species of V. cholerae-specific, virulent podoviruses that are found sporadically in cholera patient stool samples collected since 2001 in Bangladesh (Seed et al., 2011 and present study). To begin to address the geographic diversity of cholera phages in patient stools, we tested by plaque assay for the presence of phages within nine Haitian cholera patient stool samples collected in 2013. We identified one sample that had a high titer of phage (108 PFU/ml) relative to V. cholerae (105 CFU/ml). Whole genome sequencing and comparative analysis of this phage, named ICP2_2013_A_Haiti, revealed 84% identity over 93% of its genome to ICP2_2011_A, a phage isolated from Bangladesh in 2011. An alignment of the annotated ICP2 genome (Seed et al., 2011) with the three available ICP2-related isolates shows that synteny is conserved between these geographically distinct phages, with no significant genome rearrangements observed (Figure 1—figure supplement 1). The Haitian ICP2 isolate is strikingly similar, although clearly distinct, from Bangladeshi ICP2 isolates, and is the first lytic phage reported to be associated with epidemic V. cholerae in Haiti.
Remarkably, we observed that most V. cholerae isolates recovered from the same stool sample as ICP2_2013_A_Haiti were resistant to infection by this phage. We investigated phenotypic heterogeneity within this stool sample by testing 269 single colony isolates for sensitivity to ICP2_2013_A_Haiti and found that 267 (>99%) were phage-resistant. Eight phage-resistant isolates and the two phage-sensitive isolates from this sample were subjected to whole genome sequencing. All 10 isolates were isogenic except for mutations within ompU encoding the major outer membrane porin (Supplementary file 1). Furthermore, the ompU mutations in the eight phage-resistant isolates were heterogeneous. We sequenced ompU from an additional 11 phage-resistant isolates from this stool sample and found a total of six unique ompU alleles (Figure 1A). When each of these alleles was used to replace ompU in a clean genetic background, all conferred resistance to ICP2, yet produced normal amounts of OmpU (Figure 2 and data not shown), suggesting that ICP2 uses OmpU as a receptor to initiate infection and that the mutations disrupt this interaction. These data showing that intra-host V. cholerae isolates are isogenic, only differing in ICP2 resistance mutations, indicates that they diverged within the patient. Furthermore, the presence of multiple different ICP2 resistance mutations within this single host suggests that selection of phage resistance occurred multiple, independent times during the course of infection. We also found evidence for ICP2-mediated selection of OmpU mutants in isolates from Bangladesh. Analysis of the ompU sequence from 54 clinical isolates collected between 2001 and 2011 showed that 15% had non-synonomous mutations (Figure 1A, Figure 1—figure supplement 2) and these alleles were sufficient for ICP2 resistance (Figure 2). One of the OmpU mutants, G325D, was observed in clinical isolates from Haiti and Bangladesh (Figure 1A). Interestingly, seven of the eight mutant OmpU proteins detected in the Haitian and Bangladeshi clinical isolates had alterations in two predicted extracellular loops, while the eighth had an altered transmembrane segment adjacent to the first of these extracellular loops (Figure 1B). This suggests that ICP2 interacts specifically with these exposed loops to initiate infection.
We also tested for V. cholerae population heterogeneity within an ICP2-positive stool sample isolated in Bangladesh in 2011. Of 46 single colony isolates tested for sensitivity to ICP2_2011_A, 22% were resistant to the phage. Whole genome sequencing of two phage-sensitive and four phage-resistant isolates from this sample revealed that the strains were isogenic, except that phage-resistant isolates had nonsense mutations in toxR, the direct transcriptional activator of ompU (Crawford et al., 1998; Supplementary file 2). We sequenced toxR from an additional eight phage-resistant isolates from this stool sample and found that, analogous to what was observed in the Haitian patient sample, there were multiple unique toxR alleles, which included two non-synonomous and two nonsense mutations (Figure 1C). Each of the four mutant toxR alleles conferred ICP2 resistance and abrogated expression of OmpU when used to replace wild-type toxR in a clean genetic background (Figure 2). Both ICP2-sensitivity and OmpU expression were restored to the clinical toxR mutants by expressing ompU in trans or by reverting the toxR allele to wild-type (data not shown), indicating that these mutant toxR alleles are necessary and sufficient for ICP2 resistance and that this resistance is mediated through loss of OmpU expression.
To address the potential consequences of the phage-resistance mutations on V. cholerae fitness, we tested the four most frequently isolated ompU alleles and all four toxR alleles in survival and growth assays. Recent Tn-seq analyses of V. cholerae showed that OmpU is critical for fitness in an infant rabbit infection model (Fu et al., 2013; Kamp et al., 2013) and for dissemination from the host into pond water (Kamp et al., 2013). OmpU has also been shown to be important for protection against the bactericidal effect of bile salts (Provenzano et al., 2001), cationic peptides (Mathur and Waldor, 2004), and intestinal organic acids (Merrell et al., 2001). When we tested for survival in bile and for competitive fitness in pond water we found that the four OmpU mutants are fully fit (Figure 3A,B), consistent with their wild-type levels of OmpU expression (Figure 2). Moreover, the OmpU(G325D) mutant was fully virulent in a single round, competitive infection in infant rabbits (Figure 4, first column). However, the OmpU mutants, particularly the A195T and G325D mutants, showed a mild competitive defect after multiple passaging in growth medium (Figure 3C). The mild growth defect of the OmpU mutants in conjunction with the low prevalence of ICP2 (Seed et al., 2011) may explain why the ICP2-resistant ompU variants do not become fixed in the population (Figure 1—figure supplement 2). In sharp contrast, in the presence of ICP2 we observed a 10,000-fold enrichment of the OmpU(G325D) mutant over the wild-type after infection of infant rabbits (Figure 4), indicating that strong selective pressure is imposed by phage predation during V. cholerae infection. This emulates what we hypothesize happens in human infections in the presence of ICP2 and is the first demonstration of phage predation of V. cholerae in the context of a diarrheal disease model.
V. cholerae strains harboring any of the four ToxR mutations were avirulent in the infant mouse model of infection and were indistinguishable from a ΔtoxR mutant (Figure 3D). The clinical isolates containing these mutations were also at least 100-fold attenuated (Figure 3D—figure supplement 1), indicating that these isolates do not harbor compensatory mutations. The cytoplasmic amino terminus (D17-E112) of ToxR where the two non-synonomous mutations mapped has homology to the winged helix-turn-helix family of transcription activators and is involved in DNA binding and transcriptional activation (Miller et al., 1987; Ottemann et al., 1992). A number of point mutations in this domain were previously shown to abrogate activation of ompU and virulence genes (Ottemann et al., 1992; Morgan et al., 2011). The severe virulence defects of the two non-synonomous toxR mutants from the Bangladeshi patient sample (Figure 3D) are consistent with an inability of the encoded mutant ToxR proteins to activate these genes. These toxR mutants would not colonize the human small intestine and be recovered in the secretory diarrhea if ingested (Herrington et al., 1988), which further supports our conclusion that these mutants arose due to ICP2 predation during cholera infection in this individual. This also indicates that, in this patient, predation and selection of phage-resistant mutants likely occurred late in infection when colonization and virulence gene expression were no longer required for the development of acute symptoms (Merrell et al., 2002).
We have shown that lytic phages are an unexpected ‘third party’ that impose significant bactericidal pressure on an environmentally transmitted pathogen during the natural course of infection in humans. We observed that adaptations to phage predation involve tradeoffs in evolutionary fitness and provide a molecular mechanism for phage predation impacting V. cholerae transmission and seeding of environmental reservoirs. Our results highlight that host–pathogen interactions are embedded within and strongly affected by a complex microbial ecosystem. Predator-prey dynamics are notably absent between microbiotal phage and their bacterial hosts in the intestinal ecosystem in healthy humans (Reyes et al., 2010). In contrast, our findings indicate that diseased states, which are often accompanied by significant bacterial proliferation, likely facilitate these predatory interactions.
Strains utilized in this study are listed in Supplementary file 3. Stool specimens and V. cholerae isolates from the International Centre for Diarrheal Disease Research, Bangladesh were collected and stored from previous studies (Nelson et al., 2008; Seed et al., 2011, 2012, 2013). Phage (ICP2_2011_A and ICP2_2013_A_Haiti) were isolated from cholera rice-water stool samples as described (Seed et al., 2011). V. cholerae isolates were isolated from stool samples on Luria–Bertani agar with sulfamethoxazole (24 µg/ml), trimethoprim (32 μg/ml), nalidixic acid (20 µg/ml) and streptomycin (Sm) (100 µg/ml). Mutations were introduced using pCVD442-lac as previously described (Seed et al., 2012). Strains containing the pMMB67EH vector were grown in the presence of 50 μg/ml ampicillin and 1 mM isopropyl-ß-D-thiogalactopyranoside (IPTG).
V. cholerae and phage genomic libraries were generated (Lazinski and Camilli, 2013) and sequenced using an Illumina HiSeq2000 (Tufts University Core Facility). Sequence reads from each V. cholerae isolate were mapped to the reference genome V. cholerae O1 2010EL-1786 (for Haitian isolates, accession numbers NC_016445.1 and NC_016446.1) or to V. cholerae O1 N16961 (for Bangladeshi isolates, accession numbers NC_002505.1 and NC_002506.1) and mutations were identified using Varscan v2.3.6 (Koboldt et al., 2012) and CLC Genomic Workbench software (Version 6.8; CLC Bio, Denmark). Variants were called as being present (different allele than reference), absent (the same allele as reference) or low-coverage (<15×; in which case no call was made).
Cultures were grown to mid-exponential growth phase in Luria–Bertani (LB) broth at 37°C with aeration. Bacteria were re-suspended in 200 mM Tris–HCl, pH 8.0 and 2 mM EDTA. Sucrose was added to a final concentration of 20% followed by lysozyme treatment for 10 min at 37°C. Using a dry ice/ethanol bath, samples were freeze-thawed twice followed by DNAse I treament for 20 min at room temperature. Samples were spun for 5 min at 16,100×g to pellet the membrane fraction. The pellets were re-suspended in 1% Triton X-100, 10 mM MgCl2, and 50 mM Tris–HCl, pH 8.0 and incubated at 37°C for 20 min. Samples were spun again for 5 min at 16,100×g to separate the inner membrane from the outer membrane. The pellets, which contain the outer membrane proteins, were re-suspended in 200 mM Tris–HCl, pH 8.0.
Outer membrane fractions were boiled for 10 min in sample buffer containing sodium dodecyl sulfate and β-mercaptoethanol and separated on a NuPAGE 4–12% Bis-Tris polyacrylamide gel (Life Technologies, Carlsbad, CA). Protein gels were transferred to a nitrocellulose membrane for Western blotting. Membranes were probed with rabbit polyclonal antisera against V. cholerae OmpU (gift of James Kaper). A Cy5 goat anti-rabbit antibody was used to develop the blot.
Strains (grown to OD600 = 0.5) were assessed for their ability to survive 0.2% porcine bile (Sigma, St. Louis, MO) in 0.85% NaCl for 1 hr at room temperature. Pond water survival assays were done as described previously (Kamp et al., 2013) at 30°C for 48 hr. Competition experiments were performed between the strain of interest (lacZ+) and the appropriate control strain (ΔlacZ), and outputs were plated on LB agar plates containing 100 µg/ml Sm and 40 µg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal).
All animal experiments were in accordance with the rules of the Department of Laboratory Animal Medicine at Tufts Medical Center. Infant mouse colonization assays and in vitro controls were done as described previously (Seed et al., 2012). 3-day old infant rabbits were pre-treated with Cimetidine-HCL (Morton Grove Pharmaceuticals, Morton Grove, IL) 3 hr prior to infection (Kamp et al., 2013) and infected with ∼5 × 108 CFU in 2.5% sodium bicarbonate buffer (pH 9). A 1:1 mixture of wild-type to mutant (for experiments without phage) or 10:1 wild-type to mutant (for experiments including phage) was used. For experiments involving the addition of phage, phage were added immediately before intragastric inoculation to each rabbit inoculum to limit phage adsorption ex vivo (phage were in contact with the bacterial inoculum for ∼30–60 s prior to gavage of each animal). Rabbits were euthanized ∼12 hr post-inoculation and cecal and/or small intestinal fluid was collected by puncture. As with the in vitro assays, the competing strains were enumerated on LB agar plates containing 100 µg/ml Sm and 40 µg/ml X-gal.
PRED-TMBB: a web server for predicting the topology of β-barrel outer membrane proteinsNucleic Acids Research 32:W400–W404.https://doi.org/10.1093/nar/gkh417
Phages and the evolution of bacterial pathogens: from genomic rearrangements to lysogenic conversionMicrobiology and Molecular Biology Reviews 68:560–602.https://doi.org/10.1128/MMBR.68.3.560-602.2004
Regulation of virulence in Vibrio cholerae: the ToxR regulonFuture Microbiology 2:335–344.https://doi.org/10.2217/17460922.214.171.1245
Final report of the independent panel of experts on the cholera outbreak in HaitiAvailable at: http://www.un.org/News/dh/infocus/haiti/UN-cholera-report-final.pdf.
Regulatory cascade controls virulence in Vibrio choleraeProceedings of the National Academy of Sciences of USA 88:5403–5407.https://doi.org/10.1073/pnas.88.12.5403
Self-limiting nature of seasonal cholera epidemics: Role of host-mediated amplification of phageProceedings of the National Academy of Sciences of USA 102:6119–6124.https://doi.org/10.1073/pnas.0502069102
Seasonal epidemics of cholera inversely correlate with the prevalence of environmental cholera phagesProceedings of the National Academy of Sciences of USA 102:1702–1707.https://doi.org/10.1073/pnas.0408992102
Toxin, toxin-coregulated pili, and the toxR regulon are essential for Vibrio cholerae pathogenesis in humansThe Journal of Experimental Medicine 168:1487–1492.https://doi.org/10.1084/jem.168.4.1487
The Vibrio cholerae ToxR-regulated porin OmpU confers resistance to antimicrobial peptidesInfection and Immunity 72:3577–3583.https://doi.org/10.1128/IAI.72.6.3577-3583.2004
The ToxR-mediated organic acid tolerance response of Vibrio cholerae requires OmpUJournal of Bacteriology 183:2746–2754.https://doi.org/10.1128/JB.183.9.2746-2754.2001
The two faces of ToxR: activator of ompU, co-regulator of toxT in Vibrio choleraeMolecular Microbiology 81:113–128.https://doi.org/10.1111/j.1365-2958.2011.07681.x
ToxR proteins with substitutions in residues conserved with OmpR fail to activate transcription from the cholera toxin promoterJournal of Bacteriology 174:6807–6814.
Use of phoA gene fusions to identify a pilus colonization factor coordinately regulated with cholera toxinProceedings of the National Academy of Sciences of USA 84:2833–2837.https://doi.org/10.1073/pnas.84.9.2833
Pascale CossartReviewing Editor; Institut Pasteur, France
eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.
Thank you for sending your work entitled “Evolutionary Consequences of Intra-Patient Phage Predation on Microbial Populations” for consideration at eLife. Your article has been favorably evaluated by Richard Losick (Senior editor), a Reviewing editor and 2 reviewers, one of whom, Marie-Agnès Petit, has agreed to reveal her identity.
The Reviewing editor and the two reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.
Seed et al. report a very clear cut and elegant set of data providing evidence of a natural, in vivo predator prey relationship between Vibrio cholerae and ICP2-like virulent phages of the podoviridae family, upon human infection. In two patients infected with V. cholerae, and having high loads of ICP2 phage, they discovered two different panels of Vibrio mutations, in ompU and toxR, resulting in the mutation or repression of the OmpU receptor, respectively. They also establish that in a collection of 54 isolates of V. cholerae, 15% had an ompU mutation.
To our knowledge, this is the first report of a predator-prey evolution taking place in the human intestine. In addition, the toxR mutants have attenuated virulence, suggesting that the presence of the virulent phage may have contributed to the elimination of the infection by natural evolution. This is not true however for the ompU mutants that appear almost not attenuated in virulence/ survival phenotypes. The contrast between the two sets of mutants isolated is striking. As suggested by the authors, such an evolution, not reported until now for bacterial strains in healthy microbiota, may have taken place due to the disequilibrium of the resident microbiota in patients. This is a new concept, and the study certainly deserves publication.
This study also represents an extremely interesting and timely topic in light of the recent phage therapy re-birth. One of the reviewers wrote: “I was amazed by the results on ompU mutant selection by phage co-infection in these conditions and was surprised that this selective effect had never been observed in the cholerae OmpU. I went to Genbank and retrieved the various cholerae OmpU (VC0633) sequences and compared its variation to its neighboring D-alanyl-D-alanine carboxypeptidase (VC0632). I did not perform a statistical analysis, but it is obvious that OmpU is more variable than VC0632. However, this can have many different causes, but strikingly, most of the OmpU variants carry mutation(s) in the very same positions as those identified in this study (A182, VG 324/5...). A simple but thorough analysis of the OmpU mutations could add more weight to the conclusions presented in this manuscript on the selective force exerted by phage over the time.”https://doi.org/10.7554/eLife.03497.013
[…] This study also represents an extremely interesting and timely topic in light of the recent phage therapy re-birth. One of the reviewers wrote: “I was amazed by the results on ompU mutant selection by phage co-infection in these conditions and was surprised that this selective effect had never been observed in the cholerae OmpU. I went to Genbank and retrieved the various cholerae OmpU (VC0633) sequences and compared its variation to its neighboring D-alanyl-D-alanine carboxypeptidase (VC0632). I did not perform a statistical analysis, but it is obvious that OmpU is more variable than VC0632. However, this can have many different causes, but strikingly, most of the OmpU variants carry mutation(s) in the very same positions as those identified in this study (A182, VG 324/5...). A simple but thorough analysis of the OmpU mutations could add more weight to the conclusions presented in this manuscript on the selective force exerted by phage over the time.”
We thank the reviewers for their comments and idea to include more OmpU sequences. We used BLASTP to identify OmpU in other Vibrio cholerae isolates in GenBank. Homologues were aligned and WebLogo (Crooks et al., 2004) was used to generate a sequence logo in order to gain insight into variable or conserved regions. We have included this analysis in a figure below (Author response image 1).
We have also included a reference image equal to Figure 1b but with numbering of the extracellular loops to orient the reader. The mutations observed in V. cholerae isolates in this study, and experimentally validated to confer ICP2 resistance, are highlighted in yellow. The results of this analysis indicate that there is significant variability in OmpU, specifically in the predicted extracellular loops. As the reviewers point out, there are other V. cholerae isolates in GenBank that share the mutations identified in this study. There are also highly variable loops (i.e., loops 2 and 6) that we did not observe mutations in our data set, and it is not known if variability in these loops can confer phage resistance.
The strains included in the analysis below represent a broad collection of both environmental and clinical V. cholerae isolates from geographically disparate locations. Trying to identify the source of these deposited sequences is not always possible. Since OmpU represents an abundant outer membrane protein it is not surprising to observe this variability, which as the reviewers point out, is likely due to many different causes. One could imagine that the variability may be due to immune pressure, or phage pressure from ICP2 and/or other uncharacterized phages. To-date ICP2 like phages have only been isolated from clinical samples in Haiti and Bangladesh and we do not know if they co-exist with environmental non-toxigenic V. cholerae in these locations, or for that matter anywhere else in the world. We feel that the analysis provided in this manuscript which is the analysis of over 50 clinical strains in Bangladesh collected over a 10 year period, which share a niche with ICP2, is particularly relevant to making conclusions about the selective force exerted by ICP2 over time. In addition, we were able to experimentally confirm our hypothesis that these mutations confer ICP2 resistance.https://doi.org/10.7554/eLife.03497.014
- Andrew Camilli
- Andrew Camilli
- Richelle C Charles
- B Jesse Shapiro
- B Jesse Shapiro
- Kimberley D Seed
- Louise C Ivers
- Jason B Harris
- Richelle C Charles
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
The authors thank the Tufts University Core Facility for sequencing and computational support, James Kaper for providing OmpU antisera, and David Lazinski and Linc Sonenshein for commenting on the manuscript. This work was supported by US National Institutes of Health grants AI055058 (AC), AI099243 (JH/LI), AI089721 (RCC), a Massachusetts General Hospital Physician Scientist Development Award (RCC), the Natural Sciences and Engineering Research Council of Canada and the Canada Research Chairs program (BJS), and the Charles A King Trust Postdoctoral Fellowship Program (KDS). AC is a Howard Hughes Medical Institute Investigator.
Human subjects: The studies were approved by the institutional review boards of Partners HealthCare (Brigham and Women's Hospital and Massachusetts General Hospital) and Tufts University School of Medicine, and the Haitian National Ethics Committee. Written informed consent was obtained from adult participants and from guardians of children.
Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocol (#B2013-44) of Tufts University School of Medicine.
- Pascale Cossart, Institut Pasteur, France
- Received: June 4, 2014
- Accepted: July 24, 2014
- Version of Record published: August 26, 2014 (version 1)
© 2014, Seed et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Downloads (link to download the article as PDF)
Download citations (links to download the citations from this article in formats compatible with various reference manager tools)
Open citations (links to open the citations from this article in various online reference manager services)
Antimicrobial resistance (AMR) is a global threat. A better understanding of how antibiotic use and between-ward patient transfers (or connectivity) impact population-level AMR in hospital networks can help optimize antibiotic stewardship and infection control strategies. Here, we used a metapopulation framework to explain variations in the incidence of infections caused by seven major bacterial species and their drug-resistant variants in a network of 357 hospital wards. We found that ward-level antibiotic consumption volume had a stronger influence on the incidence of the more resistant pathogens, while connectivity had the most influence on hospital-endemic species and carbapenem-resistant pathogens. Piperacillin-tazobactam consumption was the strongest predictor of the cumulative incidence of infections resistant to empirical sepsis therapy. Our data provide evidence that both antibiotic use and connectivity measurably influence hospital AMR. Finally, we provide a ranking of key antibiotics by their estimated population-level impact on AMR that might help inform antimicrobial stewardship strategies.
Moving patients between wards and prescribing high levels of antibiotics increases the spread of bacterial infections that are resistant to treatment in hospitals.